Fast recognition of overlapping fruit based on maximum optimisation for apple harvesting robot

In order to improve the recognition speed and accuracy, this study aimed at overlapping apples proposes a positioning method based on the maximum optimisation, for fitting of overlapping circles. Owing to the diversity of apple growth posture, the overlap and shade between the apples will influence the recognition efficiency of apple harvesting robot vision system. Firstly, using the K-means method to segment the captured overlapping apple image under the L*a*b* colour space, lead to the overlapping apple target, and carries on the morphological processing, extracting the outline of the overlapping apple. And then to calculate the minimum distance between the pixels in the circle and the edge of outline, to obtain the minimum distance function, and find the local maximum in these functions, that is the centres of two circles. Finally, the radius is determined by the minimum distance between centres to the edge of the outline and to realise the positioning of overlapping apple, in order to verify the validity of maximum recognition method for the overlapping apples, and comparing with corrosion method and the Hough transform method, using two overlapping images captured by different apple varieties to repeat experiments. The experimental results show that, for the recognition speed and accuracy, the maximum optimisation method is optimal.